[prev in list] [next in list] [prev in thread] [next in thread] 

List:       netcf-devel
Subject:    [netcf-devel] [PATCH] Fix build errors when using	-Werror=unused-but-set-variable
From:       laine () laine ! org (Laine Stump)
Date:       2011-10-12 16:01:36
Message-ID: 4E95B9E0.60600 () laine ! org
[Download RAW message or body]

On 10/12/2011 10:46 AM, Eric Blake wrote:
> On 10/11/2011 10:20 PM, Laine Stump wrote:
>> From: Laine Stump<laine at vlap.laine.org>
>>
>> The following errors appeared when I began building on Fedora 16. In
>> all cases, the variables really are never used, so they've been
>> removed.
>>
>> +++ b/src/drv_redhat.c
>> @@ -166,11 +166,10 @@ static char *find_ifcfg_path_by_hwaddr(struct 
>> netcf *ncf, const char *mac) {
>>    * interface by checking for an entry 'DEVICE=NAME'
>>    */
>>   static char *find_ifcfg_path_by_device(struct netcf *ncf, const 
>> char *name) {
>> -    struct augeas *aug = NULL;
>>       int ndevs = 0;
>>       char **devs = NULL;
>>
>> -    aug = get_augeas(ncf);
>> +    get_augeas(ncf);
>
> I had to look, but it does indeed look like get_augeas is useful for 
> its side effects of initializing a member of ncf.  But are you risking 
> a NULL dereference if the initialization failed?  That is, should you 
> be using ERR_THROW(!get_augeas(ncf), ncf, E..., "failed to get 
> augeas") instead of blindly ignoring the return value?
>

That's what the "ERR_BAIL(ncf)" immediately following get_augeas(ncf) 
does - whenever a lower level function throws an error, it sets a flag 
in the ncf object, so on return callers can just do "ERR_BAIL(ncf)"; if 
an error has happened, ERR_BAIL() jumps to error;.

[prev in list] [next in list] [prev in thread] [next in thread]